StepFun's Step 5 Preview scores 44 on the Artificial Analysis Intelligence Index, matching Kimi K3 (max) at ~2.8x lower cost per task, but trails peers on agentic evaluations
Step 5 Preview is @StepFun_ai's new flagship model, with 600B total and 27B active parameters, succeeding Step 3.7 Flash (released May 2026). It scores 44 on the Intelligence Index, level with Kimi K3 (max) and just behind GLM-5.3 (max, 45) and Qwen3.8 Max (45)
Key takeaways:
➤ Step 5 Preview costs ~2.8x less per Intelligence Index task than models at the same score. It costs ~$0.72 per task, against ~$2.00 for Kimi K3 (max) at the same score of 44 and ~$2.01 for GLM-5.3 (max) at 45. This is driven by pricing: at $1/$2.70 per 1M input/output tokens, it is priced below both on input and output. MiMo-V2.6-Pro is the one model that scores higher (46) at a lower cost per task ($0.13)
➤ Frontier reasoning is the standout strength, and where the jump from Step 3.7 Flash is largest. Step 5 Preview scores 46% on Humanity's Last Exam, in line with Kimi K3 (max, 47%), and 21% on CritPt, between Kimi K3 (23%) and GLM-5.3 (max, 19%). Both are up sharply from Step 3.7 Flash: +25 points on HLE and +19 points on CritPt…
Step 5 Preview costs about $0.72 per Intelligence-Index task versus about $2.00 for Kimi K3 (max) at the same score of 44 and about $2.01 for GLM-5.3 (max) at 45, driven by input/output pricing of $1/$2.70 per 1M tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition →.
MiMo-V2.6-Pro is the one model that both scores higher (46) and costs less per task ($0.13) than Step 5 Preview, undercutting its cost advantage over every other peer.
Reasoning is where the jump from Step 3.7 Flash is largest: 46% on Humanity's Last Exam (+25 points) and 21% on CritPt (+19 points), landing close to Kimi K3 (max)'s 47% and 23% respectively.
On AA-Omniscience, Step 5 Preview reaches 42% accuracy (ahead of GLM-5.3 max's 34%) but attempts more questions (68% vs 55%) and hallucinates more often when it does (43% vs 30%), landing at an Omniscience Index of 16 versus GLM-5.3's 14.
It specifically lags peers on agentic evaluations: 1,566 Elo on GDPval-AA versus 1,668 (Qwen3.8 Max) and 1,646 (GLM-5.3 max), with similar gaps on Terminal-Bench 4.0 (33% vs 39%/42%), AA-Briefcase, and AutomationBench-AA.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
Why it matters
Step 5 Preview matches Kimi K3's Intelligence Index score at about 2.8x lower cost per task and shows a big reasoning jump, but trails on agentic evaluations, a reminder to check task-specific benchmarks before picking a model for AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → work.
Read the full post on X ↗
